Re: [quicwg/base-drafts] Remember UDP size limit for 0-RTT (#3498)

[Martin Thomson <[e-mail]>]

martinthomson commented on this pull request.



>  
 * initial_max_data
 * initial_max_stream_data_bidi_local
 * initial_max_stream_data_bidi_remote
 * initial_max_stream_data_uni
 * initial_max_streams_bidi
 * initial_max_streams_uni
+* max_udp_payload_size

You are suggesting then that this can work in certain narrow cases, therefore it shouldn't be a firm requirement?

The inverse of that is that is what concerns me.  Servers accept 0-RTT with reduced limits, with reasons other than the ones you describe, and all of the 0-RTT goes into the void.  Clients will routinely attempt to fill all available 0-RTT space with data.  If the limits that the server imposes matter here.

If local constraints change in a way that the server knows about, such that it will advertise a lower limit in the transport parameter, then it is better if it doesn't accept 0-RTT.

I get that this is a "no harm" scenario in the sense that the connection won't be torn down as a result of all those packets getting lost, but the result is worse than rejecting 0-RTT because the client will mark all those packets as lost as opposed to just forgetting them.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/pull/3498#discussion_r387479268